Down and Out in America: The Origins of Homelessness

SOCIOLOGy/PUBLIC POLICy/URBAN STUDIES

The most accurate and comprehensive picture of homelessness to date, this study offers a powerful explanation of its causes, proposes short- and long-term solutions,

and documents the striking contrasts between the homeless of the 1950s and 1960s and the contemporary homeless popUlation, which is younger and contains more

women, children, and blacks. 'i\n extremely important contribution to the dialogue. [Rossi's] portrait of the homeless not only fits the numbers, it matches anecdotal reality."

-William Tucker, Wall Street Journal 'i\ landmark report that documents in admirable detail the alarming scope of the problem and demonstrates its connection to the larger, more neglected, less visible

problems of extreme poverty among those the author calls the 'precariously housed. '" -John Howland, Jr., Washington Post Book World "Strongly recommended both as a guide to this social problem in general and as exemplary applied sociology." -Martin Bulmer, Times Higher Education Supplemelll

"Far and away the best study of the homeless." -Michael Novak, Forbes

PETER

H. ROSSI is Stuart A. Rice Professor of Sociology and Director of the Social

and Demographic Research Institute at the University of Massachusetts at Amherst. Cover photograph: Eugene Richards, Magnum Photos, Inc.

The University of Chicago Press www.press.uchicago.edu

ISBN 978-0-226-72829-2

1 ~lijl]lijl]

1) (

11~

,/ BORKNJ\I6763

M50

DOWN AND OUT

IN AMERICA

ZtlPADOCESKtI UNIVERZITA v Plzni Univerzitni knihovna

111111111111111111111111111111111111111111111111111111111111

DOWN AND 0 T

I

M I

The Origins of Homelessness

I Peter H. Rossi I

The University of Chicago Press Chicago and London

PETER H. ROSSI is Ihe SWart A. Rice Professor of Sociology and acting director, Social and Demogmphic Research Institute, University of Massachusetts.

The University of Chicago Press gratefully Dcknowledges the contribution of the Jensen Lectureship and the American Sociological Association toward publication of this book.

The University of Chicago Press, Chicago 6{)637 The University of Chicago Press, Ltd., London © 1989 by Peler H. Rossi AU rights reserwd. Published 1989 Printed in the United States of America

98

97 96 95

94

93

92 91

90

5432

Library of Congress Cataloging-in-Publication Data Rossi, Peter Henry, 1921Down and out in America: the origins of homelessness / Peter H. Rossi. p. em . .. Annotated bibliography of the combined homeless studies and studies of the extremely poor": p. Bibliography: p. Includes index. ISBN 0-226-72829-3 (alk, paper) I. Homelessness-United States-History. 2, Urban poor-Uniled Slates-History. l. Title.

HV4504.R67 1989 362,S'l'0973-de20

89·31598 CIP

@)

The paper used in thi.'i publication meets the minimum requirements of the American Nalional Standard for Information Sciences-Permanence of Paper for Prinled Library Materials, ANSI Z39.48-1984,

v?lZTli

For America's extremely poor and homeless

II

]1 Contents

Preface 1 2 3

ix

On the Bottom and Out on the Fringe 17

The New Homeless and the Old

Counting the Homeless and the Extremely Poor

4

Location and Economic Circumstances 5

45

The Condition of the Homeless and the Extremely Poor: 83

The Demography of Homelessness and Extreme Poverty 6 7

Vulnerability to Homelessness

143

Why We Have Homelessness and Extreme Poverty and What to Do about Them

181

Appendix A Annotated Bibliography of the Combined Homeless Studies and Studies on the Extremely Poor 213 Appendix B The Design of the Chicago Homeless Study

References Index

239 243

223

117

II

II Preface

Although I have been poor, I have never been extremely poor or homeless, for which I am thankful. As I will show in this book, neither condition is pleasant, healthful, or comfortable. To be extremely poor or homeless is to be outside the American mainstream: it is unpleasant, unhealthful, and full of pain. But lack of direct experience does not dictate an absence of either understanding or sympathy. I wrote this volume out of concern for the plight of the extremely poor and homeless of America and in the conviction that we must have good knowledge about both groups before we can devise effective means to alleviate their condition and adjust our institutions so that their numbers will eventually decline and, we hope, diminish to zero. This book had its immediate origins in a survey I designed and directed on the homeless population of Chicago, but it is rooted in a lifelong concern for those on and outside the margins. My work as a sociologist has consistently centered on the policy concerns of social programs dealing with problems of the less advantaged of our society. My most fervent wish is that this book will help those who are homeless now and prevent others from falling down and out to the fringes of our society. lowe a debt to George Orwell for paralleling the title of his insightful and sensitive Dowll and Out in London and Paris, one of the best descriptions of living close to destitution. I wish I could say I had been following in Stuart A. Rice's footsteps in studying the homeless and the extremely poor. In point of fact, at the time I chose his name for the chair I occupy at the University of Massachusetts I did not know that Rice had published several important studies of homeless men in New York during the First World War, stemming from his experience during a short term as the superintendent of New York's Municipal Lodging House. It was a great surprise to discover that the man whose name I had chosen out of admiration for his technical contributions to social research had also done important early work on homelessness. My labors on this book have been lightened by the help, kindness, and generosity of many people. First, the Robert Wood Johnson Foundation and the Pew Memorial Trust provided funds so we could undertake the Chicago Homeless Study, upon which I draw heavily in this book. The Illinois Department of Public Aid also underwrote a portion of the costs. Funding the Chi-

Preface

Preface

cago Homeless Study meant taking risks, and I admire its benefactors. It was

step, ready to listen to my problems and to offer advice that materially improved the book. There is no way I can overstate the value of the steady

not clear at the time they funded this research that my colleagues and I had the right solutions to the problems of studying a population whose members had no fixed abode. Second, I have been blessed with the best possible set of colleagues and collaborators. The staff of NORC: A Social Science Research Institute were true partners in the design of a data-collection strategy that solved the knotty problem of how to reach, count, and interview homeless persons. I am es-

pecially indebted to Dr. Martin Frankel, Dr. Mary Utne O'Brien, Dr. Sara Siegal Loevy, and Ms. Ann-Sofi Roden of the NORC staff. At the Social and Demographic Research Institute (SADRI), I was fortunate to have Professor Gene A. Fisher as a working colleague. His contributions to the measurement

of mental illness among the Chicago homeless are the basis for much of chapter 6. In addition, he undertook most of the task of preparing estimates of the size of the homeless population of Chicago. Ms. Georgianna Willis, an edvanced graduate student at SADRI, worked prodigiously to prepare the Chicago data tapes for analysis, always with great energy and care. Professor James D. Wright also made many contributions to both the design and the analysis and was always ready to read and discuss anything I wrote. Third, a special feature of this book is that it benefits greatly by drawing upon a variety of data sources; I was given access to them

an in the true spirit

of scientific collegiality. I would not have been able to enrich this work in that fashion without the extraordinary cooperation of a number of people who gave me access to their raw data. I am particularly grateful to Mr. Matthew Stagner and Dr. Harold Richman of the Chapin Hall Center for Children of the University of Chicago, who gave me copies of data from their studies of clients of Chicago General Assistance and Aid to Families with Dependent Children. Professor Charles Hoch also was gracious in providing data from his study of Chicago SRO (single-room occupancy) hotel residents. Perhaps the most unusual generosity was shown by Professor William J. Wilson of the University of Chicago, who gave me copies of data from his Chicago Urban Family Life Study before he himself had had time to analyze them. Writing this book was made far easier by a half-year sabbatical granted by the University of Massachusetts. A gracious grant from the Rockefeller Foundation made it possible for me to take a full-year sabbatical and to entice Mr. Jeffry Will as a research assistant whose skill and dogged persistence helped speed the task. I am deeply indebted to the Rockefeller Foundation's support and to Mr. Will's valuable help. A quick start on the book was made possible by my being appointed to the Jensen Lectureship in Sociology and Social Policy at Duke University. The set of lectures I prepared for presentation at Duke early in my sabbatical served

as a detailed outline for the manuscript that took me the rest of the leave to finish. My tender comrade and wife, Alice S. Rossi, helped me along at every

warmth, friendship, and intellectual excitement that surround me in her presence.

I have been helped by so many. May the virtues of this volume do honor to their kindness. Its faults are due to my inability to take full advantage of their help.

II

ONE

II

On the Bottom and Out on the Fringe

I'ortraits of the Down and Out Extreme poverty and homeless ness are useful abstractions that only dimly reflect the concrete human details of the extremely poor and the homeless. Most of this book concerns those abstractions. Some intimate knowledge

of the people described by the terms extremely poor and homeless can be gleaned from this handful of composite case histories. JOSEPH FISHER

Joseph Fisher, a black man, was thirty-three years old in 1986 when he was interviewed in a Chicago shelter. He was born in Chicago and had lived there all his life except for a two-year enlistment in the navy. His parents, now divorced, still live in that city, as do his older brother and a younger sister. Mr. Fisher did not graduate from high school, although he did qualify for a OED (general equivalency diploma) in the navy. He was employed for a short time "doing factory work" after his discharge in 1976 and has not worked steadily since. Mr. Fisher has been homeless this time for about eighteen months but has been homeless before for shorter periods. Mr. Fisher looks considerably older than his age. He is thin and seems undernourished. He rates his health as very poor and says he has problems with his kidneys and heart, both conditions that impair his ability to work. Before becoming homeless, he lived with his mother and younger sister in their apartment, but his mother asked him to leave. He would like to go back with his family but does not believe they would take him in again. He has never been married, although he fathered a child with whom he currently has no contact. He phones his mother every month or so and sometimes sleeps at her apartment.

Mr. Fisher earns ten to fifteen dollars a week hawking newspapers to drivers stopped at downtown traffic lights. Occasionally he receives some day work through a labor contractor. He was on General Assistance but

somehow lost his eligibility. Mr. Fisher would like to have steady work but no longer even tries to find regular employment. Mr. Fisher sleeps most nights in a shelter run by evangelical missionaries. lining up with several hundred others in the late afternoon to be signed in

rI On the Bottom and Out on the Fringe

Chapter One

and receive a bed assignment. In the mornings the shelter turns him Qut. Some days he can get a batch of newspapers from a circulation truck that

comes to the shelter looking for men to peddle them. Most days he spends wandering around within a few blocks of the shelter, getting lunch from a local food kitchen. Sometimes he panhandles. Mr. Fisher is an alcoholic: when he has money, he often spends it on

cheap wine, sometimes sharing his fifth with other homeless men. When he has no money, he sometimes gets drinks from friends who happen to have cash. When interviewed, he was sober, clean, and neatly dressed, though in shabby, unpressed clothing. According to the interviewer he was coherent and articulate in his answers. However, he shows the clinical signs of major depression: he sees no hope for the future, often has no appetite, and thinks about suicide several times a month. WALTER JOlIN SON

Walter Johnson was interviewed while waiting in line to receive a meal at a Chicago food kitchen set up for homeless persons. He was thirty-five at the time and lived in a room in a nearby SRO (single-room occupancy) hotel, for which he paid $200 a month. Mr. Johnson works twenty to thirty hours a week as a busboy in a downtown restaurant, earning $80 to $150 depending on the hours. His earnings allow him to pay his rent regularly but do not give him much leeway for other expenses. Once or twice a week when his cash is low, he joins the food lines. Buying a new pair of shoes forces him to show up regularly in the food lines for a week or so. He went to work in a metal fabricating plant when he finished high school and worked there more or less steadily until the plant closed four years ago. Since then the only jobs he has been able to pick up have been at slightly more than minimum wage. He has tried repeatedly to get steadier employment at higher wages. Despite high-school graduation, Mr. Johnson does not read or write with any ease. Indeed, one of his major problems is that he cannot fill out an employment application without help. He has a few friends, mostly fellow residents in his SRO. Occasionally be goes to a local bar with them. He does not drink heavily, saying he cannot

afford to. Mr. Johnson has never married. He lived with his parents until he was twenty-five but moved out when his mother died. He is not sure where his

father is now living, nor does he know the addresses of his brother and two sisters. Mr. Johnson seems very depressed about his future. He worries about

what may happen if he loses his job, fearing he might have to resort to the shelters.

PHOEBE COTT

Phoebe Cott, a white woman, was thirty-eight when she was interviewed in 1985. Born in Chicago, she fives with her mother and several younger brothers and sisters in a small apartment. She was interviewed because she is a General Assistance recipient, receiving a monthly check of $154 from the Illinois Department of Public Aid. She gives her mother $75 each month to help pay living expenses. Occasionally she works, usually as temporary help in local stores. Ms. Cott married at eighteen and was divorced after two years. She had no children. Since finishing high school, she has been employed fairly steadily, mostly at low-level clerical jobs-receptionist, retail sales clerk, data entry, and such. None of the jobs lasted long, but until 1983 she was able to rent small apartments by herself or occasionally with a live-in partner. In 1983 she was injured seriously enough in a car accident to require several weeks in the hospital and a long period of recovery. When she left the hospital, she went to live with her mother while she recovered. She has not worked steadily since. After exhausting her unemployment benefits, she applied for General Assistance and has been on the rolls for about six months. Ms. Cott hopes to find work soon. She has looked consistently, prodded somewhat by the General Assistance requirements for job search. When she finds steady work, she would like to have her own apartmen't. DENNtS CALDER

Dennis Calder was interviewed standing in line at a food kitchen in downtown Los Angeles in 1986. He was then thirty-two years old and had lived in that city since early childhood, when he arrived from Texas with his parents and siblings. Mr. Calder was married for a few years in his early twenties and is now divorced. He has not been in touch with his ex-wife or his two children for several years. Since the breakup of his marriage, he has lived by himself, mostly in rented apartments. For some of the time he lived with his parents, usually for the few weeks between his employment episodes. Hi.s parents moved back to Texas in 1980. Mr. Calder has been intermittently employed since he left high school in his junior year, working in low-paying jobs such as parking attendant, busboy, and wallboard installer. In 1980 he was convicted of breaking and entering and sentenced to a year in a minimum-security California prison. This was his fourth arrest on felony charges; the previous arrests had led to dropped charges or probation. With time off for good behavior, he returned to Los Angeles in 1981. Since his release he has had a hard time finding and keeping employment and has worked less than six months out of every year. In 1986 he was on California'S General Relief rolls, receiving $212 a month. He lived in an SRO hotel most of the time, but the benefits were not

Chapfer Dlle

On the Bottom
_t~"-.:'~."""I1!hl!!~Ll'-~IY!?Y.Ee,s,:"r."h d:S~.§[email protected] .On:nondwellings:"'-- -

IS. tatistkaTJ
~lnfiOri1e1CssneS~aJld-coliecl.mQ!!gh o~~~s o~~~~o cstlma~e \11 the flow of persons into and out of the literally homeress state.·n Unfortu- ' ilatcjy~ch of the surveys turn"CCl~Uo~expensive~nticipated, 21 and we were not able to conduct one during the summer. We did do a fall survey over a two-week period spanning I October 1985, and a two-week winter survey spanning I February 1986. The Preferred Approach to Studying the Homeless The five approaches described above vary considerably in cost and also in credibility-how far social researchers would rely on the results. Unfortucised only once. when the police escort made the decision not to enler a high-rise public housing project of national notoriety that fell into our block sample. 20. We originally planned to base some of our population estimates on a "capture/recapture" strategy, using the overlap among persons sampled at different times to estimate the total size of the population in question. Using "capturelrecapture" information from three independently drawn samples, one can also estimate with greater precision the flow into and out of the populati0l!.~_) 21. Funds were obtained through grants from the Robert Wood Johnson Foundation, the Pew Memorial Trust. and the Illinois Department of Public Aid. NORC spent approximately $450,000 on the fieldwork, and SADRI an additional $130,000 to support the direction of the study and the analysis of the resulting data. NORC expenditures included much of the development work involved in devising approaches to shelters and 10 the stratification of blocks . Given that designing a new study would be made considerably easier by experience gained in Chicago. compamblc studies could he undertaken in other cities for between $100,000 and $200,000 (in 1988 dollars).

Counting the Homeless and the Extremely Poor

Chapter Three

nately, these two characteristics are strongly correlated: the most credible approaches arc also the most expensive. In addition, within each approach there are alternative ways to proceed that affect both credibility and cost. Also, the opinions of social scientists do not necessarily rule in judging research. The least credible (to social researchers) of the national estimates of the size for the homeless population are those advanced by the National Coalition for the Homeless. Yet these estimates are the ones most cited in the mass media and by legislative bodies. '.ll\lon scientific grounds there can be no question that the modified area proba. bility strategy employed in the Chicago Homeless Study is the preferred ap1\ proach to estimating the homeless population in any jurisdictiolJJ\The strategy resolves most of the technical difficulties described earlier. To the extent that the problem of the hidden homeless can be addressed, it does so. Identification is possible through face-to-face screening. And so on. Of course no research strategy can obviate the definitional issue: about the best one can do is to be clear and explicit about the definition used. But the most important positive feature is that the estimates derived from this strategy are based on sampling theory and thus have a rational foundation. Another important feature is that the approach is general and not confined to any particular locale. (In a later section, I shall propose extending the strategy to support national estimates of the homeless.)

~

Estimating the Chicago Homeless Population One of our major goals was to estimate the average nightly prevalence of homelessness for each of the two seasons involved. Another was to estimate incidence and prevalence over a longer period. The first goal was easier to achieve, requiring only a straight projection of the average number of homeless persons found on the blocks we searched plus the count of those in the shelters. Since the fieldwork lasted about ten nights in each survey, we had to subtract any overlap that might have occurred because a person interviewed in one shelter on one night could have been found in another shelter or on the street on some other night. 21 In fact there was some overlap, mainly among shelters, as was to be expected. Using just the sample counts consisting of two sets. one for each of the tv.-o surveys, we calculated unbiased estimates of the average nightly sizes of tl1e literally homeless populations of Chicago, as shown in table 3.1. Note the much larger standard errors for the fall 1985 sample, in which we did not have a large enough sample of blocks to counter the heterogeneity of block estimates. The winter 1986 block sample is considerably larger, with correspondingly smaller standard errors. 22. Homeless persons interviewed were identified by Sodul Security numbers, names, and date of birth, infonnation obtained in the interview.

Chicago Prevalence and Incidence Estimates for 1985 and 1986 The size of a transient population such as the homeless needs to be described in a variety of ways that together are more fully descriptive. From some points of view the "point prevalence" of homelessness-the number of persons found homeless on any given night-is of greatest interest, measuring as it does the daily potential demand for shelter and other services. But other prevalence measures based on different periods may also be relevant. For example. annual prevalence measures consist of the unduplicated number of persons who are homeless over the period of a year. Measures of incidence, the number of persons who become homeless over a given period. indicate turnover in the population, especially in conjunction with the prevalence measures "discussed earlier. The two Chicago surveys formed the basis for estimates of all three kinds. Table 3.1 presents the point prevalence estimates from each of the two surveys. The two nightly prevalence estimates were close in size, 2,344 derived from the rail survey and 2,020 from the winter survey. Given the standard errors involved, it is safe to say that the number of literally homeless persons in Chicago was somewhere between 1,600 and 3,000 and most likely was about 2,300. When first released in August 1986, these average point or nightly prevalence estimates were greeted with dismay by advocates for the homeless in Chicago (Rossi 1987). In their discussions with the press and public officials, they had been fixing the size of the city's homeless popUlation between 15,000 Table 3.1 Fall 1985 and Winter 1986 Point Prevalence Estimates of the Chicago Homeless Population Survey Component

Estimate

Standard Error

A. Fall 1985 Estimates (22 September to 4 October 1985)

Shelter residents On streets or in public places Total homeless Range ± 1 SE

=

961 t,383 2,344

±13 ±735 ::!.:735

1,609-3,079

B. Winter 1986 Estimates (22 February to 7 March 1986)

Shelter residents On streets or in public places Total homeless Range ± I SE

=

1,492 528 2,020

±55 ±269 ±275

1,745-2,295

Note: Point prevalence estimates are the average duily number of persons homeless during the

relevant survey period.

Chapter Three Table 3.2

Counting rhe Homeless and the Extremely Poor

Fall and Winter Annual Chicago Homelessncss Incidence and Prevalence Estimates Estimate

Standnrd Error

A. Annual Incidence (Number of Persons Newly Homeless Each Year) Fall 1985 3 Method alpha b (conservative) Method beta~ (liberal) Winter 1986 3 Method alpha b (conservative) Method bctaC (liberal)

B. Annual Prevalence (Number of Persons Ever Homeless Fall 1985; Method alpha b (conservative) Method betae (liberal) Winter 1986 a Method alpha b (conservative) Method bela t (liberal)

5,907 2,953

:t I ,852 ±926

3,719

±475 ±479

3.752 jn

a Year) 6,962

:tl,881

4,624

±l,064

5,051 5,147

±51J

±505

aAnnual estimates were made using data on reported lengths of time currently homeless ob· rained in interviews. These estimates assume that the population is stationary and that there are no multiple entries into the homeless state during a year. bEstimales made using only data from street an'd shelter sample cases. CEstimates made using all data, including pretests and extrasample interviews.

and 25,000. When challenged on these counts, the advocates typically referred to the 1984 HUD report that reported much the same estimates. The estimates shown in table 3.1 are many magnitudes smaller than the lowest prevailing ones. Advocates were concerned that our figures would undermine their efforts to obtain more funds and support for the homeless in Chicago." Estimating annual incidence and prevalence, as shown in table 3.2, was much more difficult, requiring knowledge of the annual turnover of the Chicago homeless popUlation. From interview data on the length of each person's current episode of homeJessness, we could plot the distribution of time homeless and thus estimate the annual flows into and out of the homele" population. Data on the length of time homeless in the current episode are not the best basis for such estimates, because this episode of homelessness had not yet been completed. Furthermore, we quickly realized (but too late to change the questionnaire) that this measure was too coarse: many short-term homeless persons averaged a week or two homeless but the shortest time the measure could detcct was one month. Compensating to some degree for errors of measurement in time homeless was that the popUlation in total appeared to be stationary from survey to survey, which simplified estimates of inflow and 23. Of course. citing the t 984 HUD report was more than a little disingenuous. since it had simply printed the estimates given to HUD by the key informants interviewed in Chicago. possi· bly the advocates themselves.

outflow by constraining them to be identical. Depending on which of the several equally convincing assumptions one makes aboullhe shape of the distribution of time homeless, we arrived at a "family" of annual prevalence estimates that ranged 3,000 to 4,000 around the annual prevalence estimates shown in table 3.2. (Greater details on the prevalence estimation procedures used are given in appendix B.) Mindful that homeless persons move into and out of institutio.ns fairly frequently and that estimates based on the number of persons in shelters and on the streets necessarily omit this institutionalized component, we also tried to estimate the number of homeless persons in prisons or jails, in hospitals, or temporarily living in conventional dwellings. These estimates were also based on information gathered in the questionnaires. For example, we asked the respondents where they spent each of the seven nights before the interview, providing the basis for estimating how many persons ordinarily homeless were temporarily domiciled. The estimates for institutionalized homeless people were more shaky, resting on the proportions who reported ever being in prison or jailor hospitalized. The resulting figures are shown in table 3.3. In summary, our final estimates were as follows: Average nightly number of Chicago homeless Estimated annual number ever homeless

2,020 to 2,722 4,624 to 6,962

Estimates for Washington, D.C., and Boston Estimates based on windshield counts were conducted in two other cities (Boston Emergency Shelter Commission 1984; Robinson 1985). In Boston and the District of Columbia, observers were sent out on a systematic search of every street, with the mission to count the numbers of homeless persons encountered. The numbers present in shelters on the same nights were also obtained. In Boston on the night of 27 October 1983, a total of 2,767 persons either were in shelters or appeared homelcss on the streets and in public places. The comparable number of homeless found in the District was Table 3.3

Combined Total Point Prevalence Estimate of the Chicago Homeless Population Average of fall 1985 and winler 1986 estimales Average number of homeless dependent children Average in temporary homes~ Average institutionalized homcless h Average homeless in excluded slteltersc Estimated grand lotal of Chicago homeless

~ Based

2.182 273 42 80 145 2,722

on proportion of time thnt homeless persons interviewed claimed they spent in rented rooms or in the dwellings of others. bBnsed on hospitalization rates of the homeless. cBased on counts of available beds in shellers excluded from the sumple because they were primarily shellers for special persons other than the homeless. Includes shelters for baltered women, the physically disabled. and so on.

Chapfer Three

2,562 on the night of 31 July 1985 and was 6,454 when adjusted to take into account the homeless who had sought shelter in places not observable from the streets.24 Both these windshield surveys suffer from the limitations of that approach described earlier. It is difficult to judge whether these limitations lead to systematic uoder- or overreporting. 00 the one hand, the windshield observers may have missed persons who did not "appear" to be homeless, and on the other, some domiciled persons who looked like the homeless may have been included. In any event, both eounts are far below the highest estimates for those cities given in the HUD 1984 report, as gathered by interviewing key persons in those cities-8,000 for Boston and 10,000 for the District-and are slightly higher than the lowest estimates that HUD reported as gathered from key persons in those cities. National "Estimates" of Homelessness Almost all items in the popular press cite one estimate or another of the size of the homeless popUlation of the country. Most of the estimates can be traced to one of three sources: the 1984 HUD report on homeless ness (HUD 1984), estimates given by national spokespersons of the advocates for the homeless, and estimates prepared by Richard Freeman of the National Bureau of Economic Research (Freeman and Hall 1986). Seemingly, no other source has been bold (or foolhardy) enough to venture into the dangerous waters of national estimation. Among the citations most frequently encountered is a 1986 estimate of from 2 to 3 million homeless in the United States, whose ultimate source appears to be the Community on Creative Non-Violence. Almost as frequently cited are the considerably lower estimates produced by a Housing and Urban Development task force (HUD 1984). Somewhat less frequently encountered are references to Freeman's estimates. It is important to keep firmly in mind that there have been no direct counts of homeless persons on a national scale, accomplished by conducting either a census or a sample survey. Accordingly, every national estimate issued to date is based on local estimates, usually from key persons, employing more or less arguable assumptions about how to compile and properly "adjust" them. In the case of the most widely quoted set of estimates, it is not possible to obtain detailed information on how the calculations were made. The Commu~ nity on Creative Non-Violence has been distressingly vague in print about how it arrived at its figures. For example, in an earlier report issued by the Community (Hombs and Snyder 1982) the authors note that they based their estimates on information received from "more than 100 agencies and organizations in 25 cities and states. : .. It is as accurate an estimate as anyone in the country could offer, yet it lacks absolute statistical certainty" (p. xvi). 24. See the section on windshield surveys for il description of how this adjustment was made.

Counting the Homeless and the Extremely Poor

They go on to speculate that during 1983 the number of homeless people could reach "three million or more." Subsequent estimates have cited even larger numbers. The authors did not give any details on what information they received from the agencies and organizations and on the distribution of those agencies throughout the United States. Presumably the local figures were compiled in some way and then weighted to produce national estimates. Of course, whether such estimates arc worth any attention depends heavily on the validity of the compiling and weighting, procedures that are simply unknown. The HUD report also relies on information collected from other sources, but it provides specifics on the four approaches used for arriving at estimates. I will describe these four approaches in some detail, because each is typical of approaches used in other studies that do not rely on direct counts. The first approach involved extrapolating to the nation as a whole from the highest published estimates of the local homeless that claimed to be the total number homeless at one time. No effort was made to assess the accuracy of the estimates used. To calculate the rates of homelessness in thirty-seven cities from which published estimates were available, HUD summed the figures for all the cities involved and divided that sum by the total metropolitan popUlation in those areas. It should be noted that Rand McNally Metropolitan Areas (RMAs) were the basis for the overall popUlation figures. RMAs consist of an urban centerusually a city of 50,000 or over-plus adjacent urbanized townships and localities." Dividing the RMA population bases into the number of homeless leads to HUD's estimate of a national overall homeless rate of 0.0025 (or 25 homeless persons for each 10,000 Americans). Extrapolating from this rate to the entire nation yields a figure of 586,000 homeless. The authors of the HUD report characterize the result as an outside estimate because of the means used to calculate it. There are several reasons for believing it is an overestimate. First, published local estimates are likely to come from areas where homelessness is more severe and are less likely for cities with few homeless people. Second, the local estimates are not evaluated as to their ultimate source: the mass media show no particular tendency to publish only verifiable estimates. Finally, the procedure assumes there is a constant rate of hornelessness in cities and rural counties, which HUD argues later in the report is not so. The second approach HUD used attempted to obtain local estimates more systematically by gathering them from local experts in a national sample of 25. This is in contrast to the following alternatives that could have been IIsed by HUD: (1) standard metropolitan statistical areas, consisting of a central urban place of 50.000 or more, plus adjilcent urbanized counties. SMSAs lend to be lar£er areas than RMAs because they include

largely rural parts of the adjacent counties. (2) Central city populiltion counts, based on the assumption thal the homeless to be found in, say. BOSlon arc largely dmwn from among the domiciled within that city. The choice of RMAs as the base for HUD's calculation of homelessness mtes has been the focus of considerable criticism.

Chapter Three

Counting the Homeless and the Extremely Poor

metropolitan areas and assessing their accuracy. Because I discussed this ap-

trapolation or the extrapolation from the three studies, figures for the total

proach earlier in this chapter, I will not describe it in detail here. This keypersons approach led to an estimate of 254,000 homeless people in the nation,

homeless popUlation in 1983 range from 192,00 to 267,000. The range of estimates derived from the four approaches is 192,000 to 586,000. The BUD report (p. 19) concluded that the most reliable estimate for the number of

rural and urban areas combined.

The third approach arrived at a national estimate by extrapolating from estimates of the local homeless given by shelter operators as key persons selected in a national shelter survey. The figures gathered from the shelter operators were not assessed for accuracy but were accepted as given. By extrapolating

from these estimates, HUD came up with a national figure of 309,000 for all metropolitan areas. The 44,000 estimate for the rural homeless derived from the second approach was then added, giving an overall national estimate of 353,000, represented as the average number homeless on any given night during the winter of 1984 (December 1983 and January 1984). The fourth approach arrived at a national figure by estimating the number of homeless people in shelters and the number on the streets and summing the two. The results from the national shelter survey were used to estimate the number of homeless people found in shelters on any given day (average of 69,000). This 69,000 figure combined both metropolitan and rural sheller populations. In arriving at the number of homeless people on the streels, HUD used estimates from two sources-the 1980 census casual count and three local studies (Phoenix, Pittsburgh, and Boston)'" in which separate counts were made of shelter and street homeless populations. The casual count undertaken by the Census Bureau in 1980" was an attempt to count transients on the streets or in other public/private places where they congregate. The count took place in only some census districts; HUD extrapolated from the census figure of23,237 (based on counts covering cities comprising 12% of the United States population) to a total figure of 166,000 homeless in the nation. They adjusted this figure both upward to account for changes since the census was taken and downward to adjust for large-city bias in the casual count. The final adjusted estimate from extrapolating from the casual count was 198,000. The second method HUD used to estimate the total number of homeless on the streets nationwide was to extrapolate from three local studies of street and

shelter counts, apparently undertaken in systematic ways during mild weather when the homeless were more likely to be on the streets. HUn calculated an

average ratio of 1.78 of homeless people on the streets to those in shelters based on these three studies. Using the 69,000 figure for the homeless in shelters, HUD estimated the total number of street people at 123,000. Depending on whether the estimate for street people is taken from the casual count ex26. It is difficult to regard the Phoenix and Pittsburgh studies as on the same level of quality as the Boston windshield study. In neither Pittsburgh nor Phoenix did the researchers attempt to seck out street homeless in every possible location~ Ihey looked only in selected spots-food kitchens, shantylowns, bus slalions, and the like. 27. United SIllIes Department of Commerce, Bureau of Ihe Census (1984).

homeless people in the nation on an average night during December 1983 or

January 1984 was 250,000 to 350,000. To sum up, three of the methods-extrapolation from highest published estimates, extrapolation from estimates obtained from local interviews in a national sample of sixty metropolitan areas, and extrapolation from estimates

by a national sample of shelter operators-do not deal with actual counts. The fourth approach is the only one that estimates from actual counts of those in shelters and on the streets nationwide. As HUD points out, the sheltercount popUlation was based on a national probability sample and the street count was done by the United States Census Bureau in a nonrandom sample of cities.

Although the HUD report provided more information on how its estimates were calculated and in that sense produced considerably more credible figures than those of the Community on Creative Non-Violence, some of the neces-

sary assumptions stretch that credibility quite thin. The first BUD approach uses unevaluated estimates that simply "happen" to exist. The second approach is somewhat better in that local estimates were solicited in a systematic way, but there was no systematic assessment of the "expert opinions" used.

The third approach simply restricts the opinions solicited to a single type of source, shelter operators, whose guesses about the total number of homeless people in their cities may be better than anyone else's but are still of unknown accuracy.

Finally, the fourth approach makes the simplifying assumption that the ratio of street homeless to shelter homeless is uniform across cities and across sea-

sons. As we will see in chapter 4, the street-to-shelter ratio is highly affected by the season, although the total Chicago homeless popUlation is not. Our findings cast considerable doubt on the use of ratios from any set of cities. The final national estimates to be discussed arc those produced by Richard Freeman of the National Bureau of Economic Research (Freeman and Hall 1986). Although somewhat more sophisticated technically, Freeman's work can be regarded as an extension of the fourth HUD approach, as described above. Using somewhat different estimates of the street-to-shelter ratio, he applied those estimates to the 1983 shelter population, adjusted for an assumed growth in the latter, to come up with national estimate of 254,000 homeless persons in 1986. In a more recent unpublished paper Freeman (1988) suggests that the 1988 national totals may be as high as 500,000, given the average increases reported in the sheltered homeless counts over the years

since the 1984 HUD report. Although the national estimates discussed in this section all leave much to

be desired, they can serve as the basis for at least roughly limning out the

Chapter Three

Counting the Homeless and the Extremely Poor

magnitude of the national homeless population. The firmest figures arc those for the portion of the homeless housed nightly in emergency shelters, number-

as Manhattan, Cook, and Los Angeles-would fall inlo the sample. Any sparsely populated rural county, in contrast, would have a very small chance

ing about 100,000 in 1984. Depending on which estimate of the strcet-to-

of being selected.

shelter ratio is chosen, the street homeless are at least equal in number to the

Second, the counties so chosen would be approached using the subsampling method developed in Chicago. Lists of shellers for the homeless would be collected for each county, shelters would be sampled, and enumerations of shel-

shelter homeless and likely closer to double. Hence, based on available information and reasonable assumptions, the most believable national estimate is

that at least 300,000 people are homeless each night in this country, and possibly as many as 400,000 to 500,000 if one accepts growth rates in the past few years of between 10% and 20%. The discrepancy between the estimates of the last paragraph and the oftencited 2 to 3 million deserves some comment. Given the information available,

there are no grounds whatever for regarding the higher estimates as valid. On some level it is surely strange that guesstimates of the sort issued by local advocate groups and the Community on Creative Non-Violence should be cited repeatedly in the press as serious and valid calculations. Although one must sympathize with the journalists' strong need to have some idea of the size of the homeless popUlation and its growth trends, it is difficult to have the same feelings toward their need for ever more sensational figures.

The Case for a National Probability Survey of Homelessness There is no way to settIe how many homeless persons there may be in the United States without making a major data-gathering effort. A strategy for doing so is presented in this section. But there is also a sense in which the issue is moot-no additional evidence or greater precision is needed to establish that homelessness is a serious social problem in our country. However, the argument for a national survey of homelessness is persuasive on other

grounds. The main reason for more precise estimates is that we need the information to design appropriate ameliorative social programs. We need not only size eS-

timates but also data on the composition and location of the homeless population. Both kinds of data can be obtained in the same research operation, much as the decennial census provides information on the size of the American population and also on its composition and geographical dispersion.

The stratcgy used in the Chicago Homeless Study can be extended to pro· vide the necessary information. In rough outline, the extension would proceed

along the following lines. First, a national sample of local areas would be obtained. Using conventional area sampling methods, the strategy would call for

ter inhabitants would be undertaken, along with interviews of samples of

sheller clients. The shelter samples, properly projected, would provide unbiased estimates of the number of homeless people who use shelters. Since the objective is to make national, not local, estimates, proportionately fewer shelters would be selected in each county." The unshellered homeless would be enumerated by developing samples of small areas within each of the counties, stratified according to informed knowledge concerning the usual haunts of the street homeless. Each of the areas chosen in the sample would be searched, as in the Chicago Homeless Study, and interviews obtained from a sample of the homeless persons encountered. In Chicago we chose several hundred blocks to search: in the national survey the county subarea samples would be much smaller, although many more blocks or comparable areas would be studied all told. Finally, combining the national shdter sample and the national street sample would provide the basis for firm and precise national estimates of the homeless popUlation. In addition, interview data could be used to develop rich descriptions of the homeless. Of course this rough sketch of a national study of the homeless skips over several serious fiscal and technical obstacles. First of all, the survey would be expensive. Collecting the data necessary for drawing the subsamples of shelters and of subareas within counties would be a tedious, labor-intensive task.

Searches for the street homeless would also consume considerable resources. In current (1988) dollars, it is likely that such a survey would cost close to $10 million. Second, the homeless popUlations in small cities may not be as concentrated in locations as in a large city like Chicago, requiring technical adjustments in sampling that cannot be fully anticipated in advance. I believe the main advantages to be gained from launching a national survey effort are as follows. First, we need finn data on the size and composition of the homeless population in order to make decisions on the size and character

of the social programs necessary to ameliorate the problem of homelessness. Second, the national survey would establish a set of baseline measures against

a first-stage sampling of counties with probabilities proportionate to popula~ tion size. The resulting sample would consist of about one hundred counties,

which to assess the progress of such programs. The final goal of any national program is to reduce the prevalence of homelessness. As someone has put it, "If you don't know where you've been, you can't tell how far you've gone."

whose combined populations would mirror the population of the United States. Note that the strategy of picking counties with probability proportion-

Second, programs have to be tailored to the diversity of the homeless popula-

ate to their size usually means that all the largest urban counties-such

28. Indeed, in largely rural counties no shelters may be found at all.

Chapler Three

tion. The next few chapters provide ample evidence that the problems of the homeless are multiple, with different needs found among identifiable subgroups. Finally, without the information obtainable from a national survey, it is hard to achieve a firm intellectual understanding of the problem. Measuring Extreme Poverty In contrast to the lack of almost any estimates, good or bad, of the national homeless popUlation, poverty and unemployment are measured routinely and richly. Indeed, the monthly briefing of the Bureau of Labor Statistics at which the previous month's unemployment rate is made public provokes comments in all the daily media and often by the president and congressional leaders. Similarly, the annual release of the previous year's poverty rate is a media event that receives wide coverage and evokes considerable comment. Poverty and unemployment currently are fairly well measured. The statistical series were initially constructed by highly competent social scientists on the staffs of federal agencies and have undergone improvement after improvement since they were established in the 1940s and 1960s. The actual data collection is undertaken by the Bureau of the Census, an agency that is widely respected for both its technical competence and the absence of political partisanshi p. The unemployment rates are based on the Monthly Labor Force Survey, a quasi-longitudinal national household survey conducted by the Bureau of the Census. The survey charts the month-to-month movement of national unemployment rates and provides considerable detail on the distribution of unemployment by regions, for urban and rural areas, and by subgroups of the American popUlation. The annual March supplement to the Monthly Labor Force Survey (called the Current Population Survey) measures (among other things) how many American households fall at or below the "poverty level," a set of income boundary points established initially in 1968 (revised annually to reflect changes in basic commodity prices and in the value of the dollar) '" that divides poor from nonpoor households and families. The "poverty line" was based on estimates of the income needed to maintain households of various sizes at a minimum standard of living. Both the unemployment and the poverty indicators arc oriented to households, not individuals, being based on data obtained from large samples of households living in conventional housing units.JD Individuals living in such 29. The Monthly Labor Force Survey and the Current Population Survey are based on the same data. Income and occupation of a large sample of households (over 55,000) are gatherd every March, and the households are queried monthly throughout the year on the employment status of each adult in them. One-quarter of the households are replaced every three months, each household remaining in the sample for a full year. 30. Households and individuals living in houses or apartments or other structures ordinarily used as dwellings are covered in the sample. Persons living in "temporary quarters" such as

Counting the Homeless and the Extremely Poor

conventional housing are covered, but the surveys pass over many of the people we are concerned with here, especially those living in hotels, boardinghouses and emergency shelters. And of course homeless people hVIng on the streets 'are also not included. To be technically precise, our national unemployment and poverty statistics pertain only to that portion of the domiciled population that lives in conventional housing. . . The appropriateness of the defimtlOns used Ifi measurIng unemp:oyment and poverty have been questioned by many on other gro~nds as well. rhe Unemployment rate is defined as the percentage of persons 10 the labor force who are unemployed." Anyone not currently employed who IS not looking for work is classified as "out of the labor force," a category initially set up to account for housewives, the disabled, and retired persons. One of the controversies over the unemployment rate concerns whether "discouraged workers" should be counted as unemployed. Discouraged workers have been unemployed for long periods and have become so disheartened in their futile job searches that they no longer look for work. Under current rules such persons are classified as out of the labor force along with persons who are "keeping house" retired or disabled and therefore are not counted as unemployed." Those' questioning this classification rule claim it results in underestimations of the "true" number unemployed. In addition, the line between "looking for work" and "not looking for work" is not clear, especially when based on reports from third parties." . . . From the viewpoint of the central interests of thIS book, there IS conSIderable merit in this criticism of the current measurement of unemployment. As I will show in the next chapter, research has found that significant prop~rtions of the domiciled extremely poor and of homeless persons were not looking for work and hence are not counted among the unemployed. The measurement of poverty is even more controversial. Current measurement counts as income any earnings from employment, income transfers in monetary form (such as AFDC benefits and retirement p~nsions), an~ ot?er sources (such as rent and interest). Omitted from curr~nt Income a~e I.n-kmd transfers such as food stamps, rent subsidies, and Medicare or MedtcUld payments. Adding the cash value of the latter can shift the number .of households below the poverty level by several million (GAO 1987). Also Ifi co?troversy are the bases for assessing the minimum consumption needs of familIes; some hotels, motels, boardinghouses, or emergency shelters, and persons in institutions such as hospitals or jails are cxcluded. 3J. Keep in mind that the homeless arc omitted from all measures of unc~ployment and poverty based on either the Monthly Labor Force Survcy or the Current Populallon Survey. " 32. Exceptions are provided for persons who are temporarily on vacation or "furloughed from their regular employment. . .. . I 33. The Current Population Survey data are obtaincd from some competent mdlVlduul m cne 1 household who reports on the status of each adult member of the household. Hence, whether n mcmber is looking for work may be measured poorly.

Chapter Three

argue that since poor households pay a larger proportion of their income for housing than more prosperous families, corresponding upward adjustments

ought to be made in the income-cutoff points. Discussion of poverty in OUr country often centers on households below the

poverty line. This emphasis tends to overlook the issue of extreme poverty as defined by the households thm are a considerable distance below that cutoff point, subsuming them into a larger group that may be viewed as far better otT. Furthermore, attention is paid primarily to measuring poverty as a household characteristic. Extremely low-income individuals living in households where other adults have adequate incomes are often not counted as poor, since pre-

sumably they share in the collective fortunes of the households in which thoy are located. .

Fr~m the perspective of our concerns, the most serious problem in measur-

mg eIther unemployment or poverty is that the surveys they are based On largely overlook the homeless and those housed in hotels and rooming houses. The sample~ used are of persons living in conventional dwellings: homeless p,:ople m shelters or on the streets or domiciled persons living in hotels, mcludmg SROs, or in rooming houses or boardinghouses are simply not

counted.

Nor is the decennial census of population any better: although it does count people m shelters (and other temporary quarters), until recently the census has not made any serious attempt to enumerate (hose living outside conventional

dwellings. " Estimates of the Conventionally Domiciled Extremely Poor Despite these limitations, the Current Population Survey can be used as tbe basis for e~timating at least that portion of the extremely poor who are living In

conventIOnal homes. Bear in mind that these are estimates only of the con-

v~ntionally domiciled extremely poor and do not include the homeless, espeeIally those who are out on the streets. Nor do they include, as indicated above, persons living in hotels, motels, or rooming houses. Then: ~:e three critical assumptions in these estimates. First, parents'

responsIbIlIty for full support ends at some point in their children's early aduIt~o~~ .. Seeon?, married persons residing together have mutual support responSIbIlItIes. ThIrd, people's responsibility for their own support is modified

downwa~d when they become senior citizens. Hence the income received by an adult. m the'productive middle years who is unmarried or not living with a Spouse IS a reasonable indicator of that person's poverty (or nonpoverty)

status.

There are two critical ambiguities in these assumptions: At what age does adulthood begin? And when does Someone become a senior citizen? The esti.34. Current plans f~r the 1?90 census promise to remedy this deficiency by undertaking a spt:. clUl census of persons In public places (such as bus stations).

Counting the Homeless and thc Extremely Poor

mates given below employ two definitions of the beginning of responsible adulthood, at age twenty-two and at age thirty. The estimates also exclude students and persons living on farms, on the grounds that students are temporarily exempt from the rule that adults m.ust s~pport themselves and that farm families often act as collective economIC umts, poolIng theIr labor WIthout

parceling out family income to individual members: We also set the onset of senior citizenship at age sixty.35 Two Income defimtlOns of extreme poverty

will be used, yearly incomes under $4,000 and under $2,000, amounts that are about 76% and 38% of the poverty-level cutoffs for single persons in 1986. Table 3.4 presents the resulting estimates of extremely poor domiciled persons aged twenty-two to fifty-nine as tabulated from the 1987 Current Population Survey.:l6 Keep in mind that these data are estimates of the unattached, conventionally domiciled extremely poor, since homeless persons either in shelters or on the streets are omitted in the Current Population Survey, as are those living in hotels Or boardinghouses. Restricting the estimates in this table to unattached persons (not currently married) makes them pertain to that portion of the American population most subject to the risk of becoming literally homeless. Restricting the estimates to the nonfarm population avoids the problem of how to treat unpaid family farm workers. The only income measures available from the 1987 Current Population Survey pertained to total cash income (before taxes) received in 1986. 37 The data on marital status, student status, and living arrangements are as of March

1987. The temporal disjunction means that for some the income measures do not fairly represent current 1987 income positions; some may have become

employed in the last few months of 1986 or the first three months of 1987 and may have been currently receiving income that would place them for 1987 above the extreme poverty thresholds we used. Also, some who earned enough in 1986 were below those income thresholds in 1987. Unless these two forms of misclassification balance out to zero, the estimates presented

may be either inflated or deflated. The two columns on the right in table 3.4 pertain to the numbers and proportions of such persons whose 1986 income (before taxes) was below 35. This age cutoff was chosen to focus more clearly on persons compamble in age to the homeless and the extremely poor persons considered in earlier chapters. An estimated halfmillion unmarried persons sixty·one and over who were counted in the 1987 Current Population Survey had a 1986 income under $2,000. About half lived with relatives, most likely their adult children. 36. The Current Population Survey is a national household sample of approximately 55,000 households conducted during March of every year. Each of the sampled households is queried concerning a number of topics, including household composition, employment status, and in· come of each household member. The survey results can be weighted to project to the adult, noninstitutionalized population of the United States residing in permanent dwellings. 37. AU income from employment, pensions, unemployment benefits, and welfare programs is included. In·kind welfare benefits, such as food stamps, rent subsidies and Medicare or Medicaid, arc not included in income, nor arc subsidies from private sources such as relatives or friends.

Chapter Three Table 3.4

Counting the Homeless and the Extremely Poor

1987 Current Population Survey Data on Income

of the extremely poor are single parents, mainly women, and more arc Jiving

1986 Income Under $4,000 Living Arrangements

Number (thousands)

Under $2,000

Percentage

Number (thousands)

Percentage 18.9 4.3 0.9 2.3 26.3

15.2 33.5 11.0 11.9 73.7

Single parellfs lh'in/: with children Living alone 1,747 With parents 293 With otiter relatives 62

24.1 4.0

With nonrelulives Subtotal

141 2,243

1.9 31.3

767 173 38 93 1.071

1,229 2,348 615 802 4.994

17.0 32.4 8.5 11.1 69.0

619 1,445 447 485 2,996

0.9

Unmarried without children

Living alone With parents With other relatives With nonrelatives Subtotal

All single parellls ami childless Living alone 2,976 With parents 2,641 With otiter relmives 677 With nonrelalives 943 Total 7,237

41.0 36.4 9.3 13.0

1,386 1,618 485 578 4,067

34.1 37.8 11.9 14.2

Note: 1986 incomes of unmarried persons aged twenty-two 10 fifty-nine who arc not studenls or

living on farms.

$2,000, about the level that qualifies for General Assistance in Illinois or comparable programs in other states. According to the projections, there were more than 4 million persons in 1987 whose 1986 incomes were less than $2,000. Three out of four (74%) were unmarried persons who were not living with their children. The remaining one-fourth were single parents who were living with one or more of their own children (under eighteen). Almost two in five lived in their parents' households; the largest group (34%) consisted of unmarried persons without dependent children, and about 4% were single parents. About one in three (34%) had set up their own households; 19% were single parents in households with their children, and 15% were childless and living by themselves. The rest were in a variety of arrange-

ments, living with relatives other than their parents (12%) or sharing a dwelling with nomelatives (14%). The two columns on the left in table 3.4 arc based on a higher income cutoff, $4,000 in 1986 .. Using this more lenient definition of extreme poverty raises the number of extremely poor persons to 7.2 million, with slight shifts in their living arrangements. Under the more lenient income definition, more

in separate households. Depending on whether we take the $2,000 or $4,000 income cutoff as defining the extremely poor, there were either 4.1 million or 7.2 million unattached extremely poor persons aged twenty-two to fifty-nine in the United States in 1987. A good case can be made for either cutoff point: annual incomes under $4,000 are well below the official poverty level of $5,250 for a single person and are certainly far below the poverty level for single-parent families.'" (There were more than 2 million single parents with less than $4,000 annual income and I million with 1986 incomes under $2,000.) Table 3.4 gives a clear message. Extreme poverty affects millions of unattached adult Americans in their most productive period. The pool the literally homeless are recruited from is very large. Assuming there arc 350,000 literally homeless people in the United States and using the extreme poverty definition of $2,000 or less, about 8% of the extremely poor are literally homeless. Using the income cutoff of $4,000, the proportion drops to 5%. Given the uncertainties concerning the national estimates of the numbers of

homeless in the United States plus the laek of any information about extremely poor persons living in SROs and rooming houses, these percentages can only be regarded as estimates of magnitude. Indeed, it is reasonable to believe that the risk of becoming literally homeless among the extremely poor may be as high as one in ten. Given the incomes involved and the current prices for housing, the wonder

is that so small a proportion arc homeless. With weekly incomes ranging downward from $77 or $39, it is clear that only the very bottom of the housing market is accessible to the extremely poor. It is also obvious that the main way these extremely poor unattached adults get by is to live with their parents or with other persons. At the $2,000 annual income cutoff, about a third live on their own, and the remaining two-thirds live with their parents (mostly) or others. At the $4,000 level two out of five can be in their own quarters, and the remaining three out of five live with others." Unattached extremely poor people who arc not living with their children arc especially likely to live with others. Single-parent households, composed mostly of women and their children under eighteen, arc more likely to live separately. An argument can be made that the lower age limit in table 3.4 has been set too low. People in their twenties may still be finding their way into the labor 38. Although in many of the more generous Slates AFDC payments would bring single parents above the $4,000 income level. in less generous states, such as Te)(as, Alabama, and Mississippi, full benefits even for very large families amount to considerably less than $4,000. 39. In the Current Population Survey, nonrelatives are defined as persons who arc nol parents. children, siblings, or grandparents. Hence aunts. uncles, cousins, and other more removed kin are not included among relatives. 111Crc is no way to estimate the extent 10 which such more distant relatives figure in the housing arrangements of the very poor.

Chapter Three

Counting the Homeless and the Extremely Poor

market, and their parents and friends may be quite willing to subsidize them through periods of experimentation, with corresponding low income. Indeed, if we raise the lower age limit, the number of extremely poor persons drops

considerably. Restricting the age range to persons thirty-one to fifty-nine, the number of extremely poor persons drops to 3.8 million with 1986 incomes under $4,000 and 2.1 million with incomes under $2,000. The range of estimates produced by varying the lower age boundary and the income cutoff points are shown below:

Highest estimate: Unmarried, under $4,000, 22-59 Unmarried, under $2,000, 22-59 Lowest estimate: Unmarried, under $4,000, 31-59 Unmarried, under $2,000, 31-59

7.2 4.1 3.7 2.1

Of the estimates shown above, my own preferences are for the two that restrict the age range to twenty-two to fifty-nine. This choice is based on the assumption that this age range covers the period of the life course when most unattached adults are expected to be earning enough to sustain themselves. There can hardly be any dissent whether either $4,000 or $2,000 is a very low annual income, incapable of sustaining a reasonable standard of living for any protracted period. Whatever the choice among the four, there are clearly at least 2 million extremely poor persons and likely as many as 7 million. Of course, these are clearly also partial estimates. Since the Current Population Survey omits persons living in SROs and other "temporary" quarters, the estimates also do not refer to them. In addition, the estimates pertain only to currently unmarried persons, omitting those who are married and living with their spouses. Trends in the Number of Extremely Poor Persons

Table 3.5

1970 Current Population Survey Data on Income 1969 Total Income in 1987 Dollars Under $4,000

Living Ammgemcnts

Number (thousands)

Sillgle parents fMllg with children 667 Living alone With parents 93 21 With other relatives 24 With nonrelatives 805 Subtotal

Under S2,000

Percentage

21 3

Number (thousands)

Pcrcetllage

26

309 56 14 15 394

22

I

18 3

Ullmarried withollt chifdrl'fI Living alone With parents With other relatives With nonrelatives Subtotal

593 1,010 453 281 2,337

19 32 14 9 74

251 632 319 165 1,367

14 36 18 9 77

All sillgle parellfs and childless Living alone With parents With other relatives With nonrclatives Total

1,260

1,103 474 305 3,142

40 35 15 10

560 688 333 180 1,761

32 39 19 JO

NOle: 1969 incomes (in 1987 dollars) of unmarried persons aged twenty-two to tifty-nine who are not students or living on farms.

was no appreciable change in the proportions living with parents and living alone, but there waS an appreciable shift in living with other relatives and with nonrelatives. In 1970 more of the extremely poor lived with other relatives

Because the Current Population Survey has been taken annually, it is possible to estimate the size of the extremely poor population in earlier years and plot trends over time. Of course the estimates must take into account Ihe heavy inllation of the past two decades. Table 3.5 contains estimates of the extremely poor popUlation in 1970, using cutoff points that express 1969 in-

and fewer with nonrelatives.

come in 1987 dollar equivalents.

ling of the literally homeless popUlation just by the increase in the number of extremely poor persons in the United States. Of course there is no firm reason to make that assumption: if the stock of very inexpensive housing declined drastically in that period, that change alone could increase the risk of becom-

The most dramatic feature of table 3.5, in comparison with table 3.4, is how much smaller the 1970 extremely poor United States popUlation was: the number of extremely poor people has more than doubled since 1970, an increase of 224% both in persons earning under $4,000 and in those earning under $2,000. This increase does not simply reflect general population growth since 1970, which was less than 20%.

The message of table 3.5 is clear. The pool of extremely poor persons that the literally homeless are drawn from has increased enormously since 1970. If we assume that the proportion of literally homeless among the extremely poor has remained fairly constant over that period. then we can account for a doub-

ing homeless and also contribute to the growth of homelessness. Although the increase in the number of extremely poor people is consistent

The distribution of the extremely poor among various living arrangements

with interpretations that I will advance throughout this book, the changes in the extremely poor popUlation from 1970 to 1987 could also reflect shifts in

in 1970 appears to be quite similar to that in 1987. In 1987 the extremely poor are more likely to be single parents-31 % in 1987 versus 26% in 1970. There

their living arrangements that cannot be discerned from the Current Population Survey. Because the survey does not cover people living in hotels and

Counting the Homeless and the Extremely Poor

Chapter Three

rooming houses, a shift in the location of the extremely poor from such places to conventional dwellings would appear as an increase in that population. As I . will show in chapter 7, there has been a drastic decline in such "unconventional" dwellings in the same period, a force that, on the one hand, has produced an increase in hornelessness and, on the other, may have also increased the number of unattached pcrsons living in conventional housing, es-

pecially doubled up with other households. There are a number of reasons 10 suspect that the increase from 1970 to 1987 is not due simply to shifts in living arrangements, the strongest being that conventional dwellings are much more

expensive than the hotels and rooming houses that were demolished. In short, I believe that the 224% increase in the number of extremely poor people is a real increase in that group.

Table 3.6

1986 Incomes of Households Containing Unalluc!Jcd Persons with Incomes

under $4,000 Living Arrangements

HOU5Choid Income Under $10,000 $10,000 to $19,999 S20,000 to $29,999 $30,000 [0 $39,999 $40,000 and over

Average N

In Parents' Household

With Other

With Other

Relatives

Persons

19.9% 24.8

t5.1 [4.9 25.4 $29,851 2,348

Total

26.1%

41.0%

25.4';'0

27.4 21.4

24.8

13,9

22.9 17,5 9.2 9.5

$22,729 6t5

$18,269 803

11.2

16.6